Radeon IRC Logs For 2008-10-26

airlied_away: icewaterman: rs690 has bad swtcl, so is slow.. I think you need to turn off atmosphere in googleearth.
airlied: glisse: how did you get on?
spstarr_coding: welcome back airlied 🙂
spstarr_coding: airlied: EXA + 2D w/o kms has not crashed with your mem fixes in the DDX
spstarr_coding: i haven’t had it lock up in 2 days of many hour use
terracon: coughs
loswillios: airlied: do you know if my rs480 supports xft antialiasing in hardware?
loswillios: font antialiasing that is
icewaterman: airlied: swctl is part of the driver, right?
icewaterman: so it might get better in the future
icewaterman: airlied: thx for the hint, now googleearth works just fine
airlied: loswillios: xft font antialasing isn’t really a hw feature, however EXA should accel the operations on rs40s
airlied: rs480s
MostAwesomeDude: airlied: Back already? :3
loswillios: airlied: ah, ok. it seems EXA doesn’t do it then. A fullscreen urxvt with xft font renders very slowly on ps auxf for example
airlied: MostAwesomeDude: yup 2.5 days on the beach 🙂
airlied: loswillios: do oprofile or sysprof.
airlied: loswillios: make sure it says render accel is enable.
loswillios: ok
loswillios: (II) RADEON(0): Render acceleration enabled for R300/R400/R500 type cards.
loswillios: pardon my ignorance but how would I do oprofile?
airlied: loswillios: so it “should” work maybe enable fallback printgin in the -ati driver
airlied: src/radeon_exa_render.c
airlied: loswillios: oprofile needs debug symbols for things installed, oprofile.sf.net cotains more info
airlied: but fallback tracing might at least see the issue
loswillios: thanks
loswillios: is it just the -ati driver which I need to install with added symbols?
loswillios: hrm, a ./configure doesn’t show anything about symbols/debug
airlied: loswillios: if you build from src it’ll be fine.
airlied: you might need server symbols for EXA things.
airlied: but I’d try the fallback stuff first its much easier to play with
loswillios: ok. I can’t figure out where in src/radeon_exa_render.c fallback printing should be enabled. could you give me a pointer
airlied: oops.. src/radeno_exa.c
airlied: RADEON_TRACE_FALL
loswillios: ok, up and running. thanks
loswillios: airlied: _lots_ of R300CheckCompositeTexture: Unsupported picture format 0x1011000
loswillios: 3,6M 26. Okt 09:33 /var/log/Xorg.0.log :p
loswillios: and this in syslog: http://pastebin.com/m2c0e4b1a
loswillios: should I open a bugreport?
airlied: ah its one 1bpp fonts..
loswillios: URxvt.font: xft:Terminus
airlied: we don’t accel 1bpp fonts, they aren’t truetype
airlied: at least thats my understanding..
icewaterman: airlied: btw. this this also helped with my intel graphics problem on google earth
airlied: it might be possible the glyph cache in the master server…
airlied: icewaterman: yup its CPU bound.. doing f^n for vertex shading
airlied: oops doing e^x
icewaterman: airlied: so it is a principle problem and not a problem with the driver
airlied: icewaterman: pretty much a mesa problem, the hw has no vertex shaders
loswillios: airlied: so I would need a truetype font (which Terminus apparently isn’t)?
loswillios: apparently there’s a Terminus.ttf available
icewaterman: uhm btw. any developer on this driver interested in an x850xt?
icewaterman: i am willing to donate it to the project, though shipping costs outside germany will be too expensive, so i will only cover for the 4.20€ the shipping in germany costs if it is supposed to be sent somewhere else.
glisse: airlied: well after doing the code i think it’s easier to do a table of ptr for cs
glisse: so we can grow reloc or ib array independantly without complexe code
glisse: otherwise i am just having fun with all area in r300 & radeon which assume sarea existence
rx__: 🙂
glisse: airlied: would be nice to agree on a place were to put modesetting work, like your linux tree
paxcoder: radeon keeps freezing my machine once in a while when I minimize a window. can’t even restart X or change to tty, i have no choice but to restart the machine. I’ve got Radeon X1500. And it’s not mesa dri, I know because I hadn’t have it (or had an older version) and it still crashed. When I don’t use radeon it doesn’t crash.
paxcoder: *doesn’t freeze
paxcoder: just so you know.
icewaterman: paxcoder: what are your xorg settings for the radeon drivers?
icewaterman: i only use: Option “AccelDFS” “on”
icewaterman: Option “AccelMethod” “EXA”
airlied: glisse: yup, we could probably do with a shared area for working on it, I’ve mostly be using modesetting-gem, its just since I have to fix bugs in Fedora I generally fix that tree first.
airlied: zzzz &
paxcoder: icewaterman: wait, I’ll upload my xorg…
paxcoder: icewaterman, here: http://pastebin.com/d58b9b163
icewaterman: paxcoder: the only different settings i use are: http://pastebin.com/m1ba3994f
icewaterman: or lets say the only relevant settings that are different to yours
paxcoder: !dfs
paxcoder: icewaterman, what’s that?
paxcoder: !exa too
icewaterman: http://pastebin.com/m799f194f
icewaterman: exa is an alternative method to default XAA
icewaterman: see man radeon for details
icewaterman: you probably need to set accelmethod to exa only, either because you are on agp and acceldfs is unsafe or because you are using pcie and it is default 🙂
paxcoder: icewaterman, i have embeded graphics
paxcoder: icewaterman?
paxcoder: nothing…
icewaterman: paxcoder: same here
icewaterman: onboard graphics card with 128mb shared memory
icewaterman: on windows it is not that slow for older games like max payne (ok, i admit, that game is >7 years old), but still fun
paxcoder: icewaterman, will try what you gave me. although I won’t get my hopes up… thx
icewaterman: paxcoder: exa works better than xaa on newer cards afaik
paxcoder: kk
defenderBG: hi guys
defenderBG: today at 4 o’clock in the morning i was talking with a few guys about changing from fglrx to OSS ati drivers
defenderBG: after quite a lot of work we got to the point where i get this error: undefined symbol: __driCreateNewScreen_20050727
defenderBG: (EE) AIGLX error: dlsym for __driCreateNewScreen_20050727 failed (/usr/lib64/dri/r300_dri.so: undefined symbol: __driCreateNewScreen_20050727)
defenderBG: does anyone know how to fix it
mikkoc: newer xserver i think
defenderBG: I am using gentoo, amd64
mikkoc: which version do you have?
defenderBG: my x server is 1.4.2, should i go 1.5.2?
mikkoc: yes
defenderBG: thanks mikkoc, see you in 40 minutes 😉
defenderBG: mikkoc: i have another question, from http://www.free3d.org/ i can see that some guys with the same video card used: ColorTiling=on,EnablePageFlip=On
defenderBG: how do I set this and where?
mikkoc: my guess would be xorg.conf… but don’t know where exactly
mikkoc: probably the driver section tho
defenderBG: do you meen the device section where the driver is ati?
defenderBG: i am not good with xorg.conf
mikkoc: yea
defenderBG: ok, thanks
icewaterman: defenderBG: man radeon
defenderBG: icewaterman: what for?
icewaterman: the settings
icewaterman: where to put and what they mean is what you can find in the manpage
defenderBG: yep, i figured that out already
defenderBG: thanks
icewaterman: np
tymonn: hi there
tymonn: is it possible to install a radeon mobility HD2400 yet?
adamk: The radeon driver will work with such a GPU, but there is no acceleration yet.
tymonn: adamk: when do you think it will be ok?
adamk: No idea.
icewaterman: tymonn: typical answer for that question: when it’s done
tymonn: icewaterman: sometimes there’s a release date…..
icewaterman: michelangelo did say that to the pope when painting his famous picture at the Sistine Chapel
icewaterman: tymonn: with driver programming there is rarely a release date, especially if development is not pushed by commercial interests alone
tymonn: indeed icewaterman 🙂
tymonn: icewaterman: what driver are you using?
icewaterman: tymonn: the one from ubuntu 8.10
icewaterman: 6.9.0+git20081003
icewaterman: but i have an onboard graphics card (x1200)
tymonn: ubuntu? blergh 😛
tymonn: debian <3
icewaterman: tymonn: if you want to wait until debian includes the proper driver for your card into stable branch you can as well trash your card now
icewaterman: i like debian, but only the stable branch, all other have either high update frequency and often break the packages all over the approx 2 years, but stable is only suitable for servers or desktops where you do not rely on current software or drivers.
icewaterman: but i’ve had this discussion many times, it is both off-topic and pointless
tymonn: icewaterman: ubuntu only does large updates every 6 months
icewaterman: ubuntu is in a way very nice, because it is as current as testing in debian, but you only encounter a lot of broken packages in the beginning, then you have a stable system for half a year until the next upgrade
tymonn: debian is always on the bleeding edge
icewaterman: tymonn: that only holds for testing
icewaterman: or even unstable – which is much wose
icewaterman: worse
tymonn: i used unstable for 6 months and it worked always great
tymonn: im using testing now just for the lulz
tymonn: becaz unstable is pretty stable
icewaterman: tymonn: i used unstable for <3 weeks and it screwed my box twice entirely
tymonn: “experimental” on the other hand. oh boy 😀
icewaterman: once was because the cryptsetup package was broken – destroyed almost all my files
icewaterman: and i do not remember what the other one was
icewaterman: luckily i had a backup
icewaterman: for the data only, so i had to reinstall once, the second time i didnt even bother choosing unstable
rmh3093: airlied, ping
rmh3093: is there anyone here that cares about exa and kms not playing nicely together on my box?
terracon: airlied any R300 fixes going in today?
terracon: hmm 2 spstarr’s. One at work and one at desk, Buuuut. Shouldn’t you be at your desk working?
terracon: maybe it should be spstarr_sleep
adamk_: Trying to play ut2004 with an x1950, and current Xorg/Mesa, I get this message ‘Mesa 7.3-devel implementation error: radeon_program_pair.c::allocate_input_registers(): Don’t know how to handle inputs 0x8’
adamk: Oh, google is giving me this: http://www.mail-archive.com/[email protected]/msg36457.html
adamk: I guess it’s just a waiting game now for it to be fixed.
otaylor: adamk: there was some work in the last week going on with fog, don’t know the details
terracon: maybe in 2010 ut2004 will work correctly. 6 years in progress, not bad
adamk: It used to work just fine.
terracon: I can’t run it in linux due to the 50% performance loss compared to windows
adamk: I never played it into windows and always found the performance in linux just fine.
icewaterman: does the open source driver allow to use the card for number crunching?
khetzal: hello
khetzal: i’ve a problem with my radeon driver
khetzal: i’m using gentoo and when I start X with no dri, it work very good
khetzal: but when i activate dri, X use 99% cpu, all freeze and I can’t kill X
khetzal: i’m using an ati mobility 9000
khetzal: is someone know this bug ?
khetzal: http://pastebin.com/m7cc2087b here there is my logs
logari81: khetzal: have you tried EXA instead of XAA?
khetzal: i will test now
khetzal: i’ve the same freeze
khetzal: but it’s strange, X start, i can see the mouse and directly after it freeze, I can mouve the mouse on the screen but i can’t kill X or go back to a tty
icewaterman: ok, i got myself a silent x1600pro, so i have better performance than with my onboard card
khetzal: (i’ve a laptop)
icewaterman: khetzal: your problem happened to me as well, when i set some options for radeon.
icewaterman: try setting accelmethod to exa but skip all other settings for the card
khetzal: but i’ve no special option :/
khetzal: http://pastebin.com/d64b42b69
khetzal: this is my xorg.conf
khetzal: i’ve tried with Load “GLcore” and it doesn’t work
khetzal: and with BusID “PCI:1:0:0”
icewaterman: busid only matters if you have multiple cards
icewaterman: khetzal: in general those old cards worked fine with older versions of the driver
icewaterman: seems to be a regression
icewaterman: khetzal: dunno about the driver: http://pastebin.com/m5eaf3548 <- try replacing your module section with this one
khetzal: ok icewaterman i will try now
khetzal: with a live cd of backtrack, my card work great with acceleration :/
khetzal: icewaterman, with your module section, my X launch correctly, but i’ve not any acceleration
khetzal: (like when i start X without dri)
khetzal: if i add load glx, all work
khetzal: but if i all load dri
khetzal: it freeze
icewaterman: those modules are loaded automatically for me
khetzal: not for me :/
icewaterman: which is why i do not load them in xorg.conf
khetzal: i will compile a 2.6.16 kernel to see if it’s a regression from the dri / radeon
icewaterman: khetzal: could be but dont expect that old a driver to work with recent xorg drivers
icewaterman: they come with different microcode
khetzal: :/
icewaterman: which could be the problem
khetzal: a problem between the gentoo’s xorg and the kernel ?
khetzal: yes it’s possible, i’ve to try with the original gentoo’s kernel
icewaterman: which version of xorg is it?
khetzal: how to see the xorg version ?
icewaterman: its been a while since i used gentoo, but use your package manager to find out
logari81: Xorg -version
icewaterman: or that
khetzal: X Window System Version 1.3.0
khetzal: and x11 7.2
icewaterman: khetzal: maybe 2.6.16 will do indeed
yangman: http://wiki.x.org/wiki/Radeon%20ASICs
yangman: well, that took a while…
khetzal: hey, xorg-x11 il marked as not installed, is it normal ?
khetzal: my card is not in this list
khetzal: it’s r250
khetzal: 900
khetzal: 9000
yangman: khetzal: right. the source I used doesn’t contain ASICs lower than r300. the table needs to be supplemented
icewaterman: yangman: there is already a table for those on dri.org
icewaterman: http://dri.freedesktop.org/wiki/ATIRadeon
yangman: so there is
rmh3093: glisse, ping
glisse: rmh3093: pong
rmh3093: i saw a xf86-video-kms under your name, whats that about, is it still alive?
rx__: that’s old 😛
glisse: rmh3093: it’s was a shot at a unified ddx for kms were each driver (intel,nouveau,radeon) would have a submodule
glisse: but it could now be considered as dead
rmh3093: ic
glisse: i think eventually we will grow kms support directly into the xserver
airlied: glisse: hey…
glisse: airlied: so i was playing with dri2, and i am starting to rework nha bufmgr and do a common stuff into libdrm (like libdrm_intel)
glisse: also cleaning up mesa assumption of sarea existence, by abstracting informations sarea does provide
airlied: glisse: oh nice.. do you think we can move r300_mem.c into libdrm, or are you just doing GEM?
airlied: rmh3093: radeonhd +kms isn’t going to work.
airlied: there isn’t any point in running radeonhd on top of kms.
glisse: airlied: just doing gem but r300_mem will stay in mesa
glisse: only the gem implementation is factor out for use by ddx too
airlied: glisse: cool thats what I thought made sense 🙂
rmh3093: airlied, im trying to upgrade to xorg git
rmh3093: so i can get dri2
airlied: how are you handling the reloc emission into the command stream?
glisse: airlied: my hope is that in one year we could do rm r300_mem or even rm dri1* 🙂
airlied: rmh3093: radeon doesn’t do dri2 yet..
rmh3093: and then i will use that radeon driver in your special repo
airlied: glisse: so you return the ID to emit into the command stream or something?
airlied: glisse: or handle cmdbuf inside bumgr?
airlied: bufmgr even.
rmh3093: airlied, so radeon-gem-cs isnt going to work?
glisse: well i do have radeon_bo and radeon_cs and radeon_cs depend on radeon_bo interface
airlied: rmh3093: it should work fine with any X server > 1.5
rmh3093: ok yeah is what im trying to get working now
rmh3093: but it is looking for dri2.h
rmh3093: which i didnt have
rmh3093: so now its xorg updatetime
rmh3093: but i had 1.5.2
rmh3093: hmm
airlied: –disable-dri2 will work..
airlied: glisse: so radeon_cs is in libdrm?
airlied: glisse: btw you went with array of chunks or linked list of chunks?
rmh3093: yeah i just saw that in the ebuild
glisse: airlied: radeon_cs is in libdrm-radeon aslong as with gem-cs ioctl code for it
glisse: it will look a lot like nha things just reloc which are differents
glisse: for cs i think the best way is an array of chunk ptr ie array of struct :
glisse: chunk{int type; int ndwords; void*chunk;}
glisse: so kernel will only alloc the IB chunk and copy from it
spstarr: hmm
airlied: cool.. sounds like writev….
glisse: the in ib packet3(nop,offsetindwords_into_reloc_chunk)
glisse: after each packet needing a reloc
glisse: so basic cs parsing is the same
glisse: only lookup is different
glisse: also i enforce having reloc_chunk to have one reference only for each bo
airlied: glisse: do you want offset in dwords? or offset in reloc_size?
airlied: offset in reloc_size might allow changing reloc sizes later easier.
glisse: offset in dwords so we can grow reloc chunk
glisse: and enforcing relocstruct to be dword aligned
airlied: well offset in size lets us do that as well, the lookup function just needs to change.
glisse: well could be size, i just though that restricting to dword was fine 🙂
airlied: is the reloc format okay? or do we want to change it btw?
glisse: i change a bit the reloc somethings like that: struct reloc(uint32 handle, uint32 startoffset, uint32 size, uint32 padforeasierkernellife)
glisse: the idea of startoffset and size is to tell the kernel how much of the bo is needed
glisse: so kernel can optimize and only validate necessary pages
glisse: of course userspace have to ensure that if the same bo is used several time the startoffset and size cover the proper area
airlied: glisse: where are the domains?
glisse: oups forget the domains 🙂
glisse: but they are their
airlied: be careful with startoffset, need to check for bo overflows and wrapping 🙂
glisse: airlied: yup that’s kernel tricky part
glisse: but its pretty easy to do
airlied: yeah just check against bo size and check for wrapping.
glisse: i think i will also add bo size checking like checking that coloroffset is bigenough for in regards to cliprect setup
glisse: color&back buffer are the hardest to check
rmh3093: airlied, are you still maintaining that radeon-gem-cs branch, its looking for dri_sarea, but the r300-bufmgr mesa branch has sarea.h
airlied: rmh3093: ~airlied/radeon-gem-cs is what we ship in Fedora.
airlied: it shouldn’t need anything from mesa to build against.
rmh3093: errr, my fault its xorg-server thats looking for the wrong file
rmh3093: MostAwesomeDude, hey
MostAwesomeDude: rmh3093: Yo.
rmh3093: whats new
MostAwesomeDude: Nothin’.
rmh3093: MostAwesomeDude, i have a 1680×1050 fb finally 😉
rmh3093: feels so good
MostAwesomeDude: Yeah, I know.
rmh3093: MostAwesomeDude, i just cant get an xorg with dri2 to install
rmh3093: :/
MostAwesomeDude: Hm.
spstarr: DRI2 isn’t useful for radeon yet 🙂
rmh3093: well i cant build airlied’s driver with out it
rmh3093: its complaing i dont have the right headers
spstarr: ?
spstarr: I didn’t know airlied had DRI2 work started yet
rmh3093: actually i think im getting so mixed up with all the crap im building
rmh3093: seems like his driver is isnt what was complaining about dri2
rmh3093: at least not atm
rmh3093: ‘struct ’ has no member named ‘drmmode’
airlied: rmh3093: you need libdrm from modesetting-gem branch
airlied: not the kernel stuff just the libdrm.
rmh3093: ahhh
rmh3093: thats what im missing
rmh3093: thanks
airlied: rmh3093: airlied.livejournal.com
airlied: contains the howto.
rmh3093: i saw that
rmh3093: was following it
rmh3093: must missed that step
airlied: glisse: any patches I could hack on :), I’d like to try and make Fedora support close to upstream APIs before tomorrow devel freeze 🙂
glisse: airlied: well i still haven’t done all the code
glisse: i mostly got header file and bit of c file
glisse: but while writing i am tweaking things here and their
glisse: i don’t think i can provide anysolid code ie working before this week 🙂
spstarr: airlied: any stuff for me to test?
airlied: spstarr: not really, was away..
airlied: glisse: cool.. will try and fix the bugs in my current code 🙂
spstarr: airlied: if freeze is tomorrow, perhaps disable kms for r3xx 🙁
airlied: spstarr: devel freeze..
airlied: spstarr: I’ll disable for AGP only.
airlied: my r300s all work now under PCIE.
spstarr: ok
soopurman: For what it’s worth, I’ve got a problem I’m working on, or would like to try to get working on more accurately, that I’d love to see addressed in Fedora. Not sure if it’s too late to make any progress before tomorrow though. Basically I have an X800 AGP (R420 “JH”) that I can’t get working on PowerPC. I can describe the problem in detail for you if you think this is something you might have time to help me work on…?
marcheu: try changing the card slot 🙂
soopurman: …sorry, i guess i don’t get the joke…
marcheu: it’s not a joke, try changing the slot the card is in
marcheu: i.e. move the card
marcheu: hmm nvm it’s no a PCIE…
soopurman: if i don’t use Option “NoAccel” then it nearly freezes the machine completely (the console is completely unresponsive, and both an ssh login and a serial terminal are too slow to be even barely usable). If I do use Option “NoAccel” then it gets confused about which plug is connected to the monitor and defaults to a basic 1024×768 resolution, then when I restart X (logout) it hangs the console again.
soopurman: I tried compiling a fresh git-clone of xf86-video-ati but I must not have the right development packages or something because it chokes on the “__user” tokens in radeon_drm.h
soopurman: this is on a clean install of Fedora 10 test snapshot 3
sini: hey guys I’ve gotten DRI working with xorg 7.4 & the latest radeon driver under gentoo with a r516 based card… but now I’ve installed debian and am having trouble with dri.
rmh3093: airlied, i cant start X using that driver of your, i get a GART error
rmh3093: (EE) RADEON(0): [gart] GART failed to initialize. Disabling the DRI.
rmh3093: airlied, is there any other info you might need
airlied: rmh3093: wierd.. you sure it included modesetting.. have a look at the configure output.. check it found xf86drmMode.h
rmh3093: Redman xf86-video-ati # ./configure –prefix=/usr | grep xf86drmMode
rmh3093: checking for xf86drmMode.h… yes
sini: I have no idea why I don’t have DRI. :-/
rmh3093: airlied, yeah i found it
airlied: rmh3093: hmm.. can yuou pastebin Xorg log file?
rmh3093: airlied, http://rafb.net/p/Np38TK69.html
airlied: wierd.. looks like a problem with kms, did you modprobe fbcon before or after radeon?
rmh3093: airlied, its all built into the kernel
rmh3093: airlied, so i have no control over that
airlied: rmh3093: did you reboot since you started the other drivers?
rmh3093: no
airlied: try that…
rmh3093: brb
airlied: I’d rmmod radeon but if its built in….
sini: I don’t understand why DRI doesn’t work 🙁
airlied: pastebin a logfile..
sini: one sec doing just that
sini: http://pastebin.ca/1237287
rmh3093: airlied, success finally, i have kms and decent exa performance 😀
airlied: sini: LIBGL_DEBUG=verbose glxinfo
airlied: sini: and remove AgpFastWrites option
airlied: rmh3093: excellent.
rmh3093: airlied, so before i switched to kms, my fps were 2800 and now they are about 300, i know its not a benchmark, but that is just cause the drivers early stage right
rmh3093: there isnt something else going on right
airlied: rmh3093: you are probably getting the wrong mesa branch.
sini: removed the option one sec
airlied: rmh3093: there isn’t really a final r300 mesa driver for kms yet.
airlied: its just a pile of hacks…
rmh3093: airlied, im using the r300-buffer one
airlied: http://cgit.freedesktop.org/~airlied/mesa/log/?h=r300-bufmgr-fedora
airlied: is what I’m shipping at the moment..
airlied: but its a pile of hacks so not suitable for mainline yet.
rmh3093: i dont care i just want toplay, can ipull from that address?
airlied: yup should be useable..
airlied: git://people.freedesktop.org/~airlied/mesa r300-bufmgr-fedora branch
sini: http://pastebin.ca/1237291 that’s glxinfo
airlied: sini: you need a newer mesa 🙂
sini: I’m already running debian unstable. 😐
airlied: has it goes Mesa 7.1 or 7.2?
chithead: sini: debian unstable has only mesa 7.0 because they are in freeze
sini: stupid lenny 😛
sini: do I need to install some stuff from experimental or just wait it out. 😐
airlied: its in experimental I assume.
chithead: mesa 7.2 is in debian experimental. I don’t expect debian to put mesa 7.2 in unstable before the next release
spstarr: airlied: I guess it will take a little more time for kms AGP support to be ready, but I’ll help test whenever you have stuff
airlied: spstarr: I’ll probably give up on AGP 🙂
spstarr: nooo
spstarr: you’re so close
airlied: like its always required hand tuning even with X.org
airlied: to get the mode from crashing things.
spstarr: airlied: it does work for me, its just the busy wait issues now
airlied: spstarr: okay well that isn’t AGPs fault..
spstarr: right
airlied: terracon has AGP issues
spstarr: the r300 itself i was having issues though
rmh3093: airlied, would i have to rebuild xorg-server and rebuild mesa, cause i just rebuild mesa and the ati driver and 3d is now horrible w/ exa
spstarr: airlied: perhaps agd5f can help with that?
airlied: rmh3093: just mesa… are you getting dri?
airlied: spstarr: help with what?
spstarr: airlied: the agp issues
rmh3093: airlied, well i have it enabled but idont know if im _getting_ it
airlied: spstarr: nothing to be done, its mostly chips issues…
airlied: rmh3093: look at glxinfo
spstarr: ugh
airlied: chipset issues.
airlied: thats why AGP is always crappy…
airlied: it might be possible to limit based on known broken combinations.
airlied: so we can quirk the driver if we know the chipset/gpu combo is bad.
rmh3093: airlied, $ glxinfo -t | grep dri i get nothing
airlied: we have some patches for the DDX from Ubuntu just need to have a look.
spstarr: that should be possible to do yeah
airlied: rmh3093: pastebin LIBGL_DEBUG=verbose glxinfo
rmh3093: airlied, http://rafb.net/p/GTV0MZ33.html
rmh3093: airlied, oops
rmh3093: that was wrong
rmh3093: airlied, http://rafb.net/p/ixdlDd75.html
rmh3093: there thats better
defenderBG: hi all
airlied: rmh3093: so you have no dri
defenderBG: i managed to get everything working exect… if i am not su i cannot use direct rendering
airlied: its using swrast.
airlied: pastebing xorg log
defenderBG: how can i repair this problem
rmh3093: airlied, http://rafb.net/p/9sb2N650.html
airlied: defenderBG: chmod a+rw /dev/dri/card0 temporarily fixes it.
airlied: rmh3093: that driver isn’t using kms
defenderBG: airlied: and if i want to fix it forever (/dev is respawned every restart) how do i do that?
rmh3093: hmmm wtf
airlied: defenderBG: depends on the distro how to fix it permanently
defenderBG: airlied: gentoo…?
airlied: defenderBG: need to ask a gentoo person, there used to be a section in the xorg.conf, but udev can also do it now. or pam_console
defenderBG: airlied: glxinfo | grep direct direct rendering: No (If you want to find out why, try setting LIBGL_DEBUG=verbose)
defenderBG: airlied: I have Section “DRI” Mode 0666 EndSection
airlied: defenderBG: then its udev or pam_console or whatever gentoo uses to change persmission when you login.
defenderBG: airlied: thanks, will ask in #gentoo
rmh3093: airlied, dri is finally working
rmh3093: thanks for your help
terracon: mostawesomenessdude
EruditeHermit: airlied: I just tried snapshot 3 livecd and I was able to boot fc10 for the first time!
EruditeHermit: airlied: nice work
spstarr: EruditeHermit: which chip
EruditeHermit: rv350
spstarr: EruditeHermit: it sorta works, just dont use composite
EruditeHermit: it even detected my DPI and passed it to gnome
spstarr: EruditeHermit: there are some busy wait problems still
EruditeHermit: no distro has ever done that
EruditeHermit: spstarr: I noticed some corruption in composite
spstarr: yep
EruditeHermit: the impressive thing is the DPI detection
EruditeHermit: usually X detects DPI fine, but gnome always defaults to 96
EruditeHermit: but fc10 passed X dpi to gnome
EruditeHermit: spstarr: is it the same 3D driver for the mem manager branch?
spstarr: i think so
EruditeHermit: so it doesn’t yet take advantage of the memory manager
airlied: EruditeHermit: 3D doesn’t use the memory manager properly yet… it just sits on top of static allocations.
EruditeHermit: but if its the same 3D code, shouldn’t it work the same for composite?
airlied: it does work fine, but it can’t do zero copy texture from pixmap
airlied: for me compiz works fine, just a bit slower than it could be
EruditeHermit: I get corruptions that I don’t on my ubuntu box
EruditeHermit: it disappears when I redraw on that part of the screen
EruditeHermit: by moving a window over it
EruditeHermit: etc
EruditeHermit: airlied: I saw the boot process on phoronix site and it looked amazing
EruditeHermit: however, I don’t get the same boot on my box
EruditeHermit: I get a progress bar on a black background. I don’t get the fedora splash walpaper with spinning stuff
EruditeHermit: does that only work for a HD install?
airlied: EruditeHermit: should work for everything
EruditeHermit: odd
airlied: have you got rhgb on the kernel command line?
EruditeHermit: oh, I need to specify?
EruditeHermit: its not the default boot option?
EruditeHermit: oh wait isn’t rhgb the old one?
airlied: it shuold be the default.
EruditeHermit: which is the default?
airlied: the option is still called rhgb.
EruditeHermit: oh
EruditeHermit: hmm, I just made a USB live image from the ISO
EruditeHermit: using fedora liveusb-creator
EruditeHermit: and booted that using default options
airlied: oh that should work
airlied: oh maybe the kernel might not load the correct modules on time.
EruditeHermit: what I see is I get a black background and the progress bar
airlied: is the screen in graphical or text mode?
EruditeHermit: 1/2 way through it goes from low resolution to higher resolution and the progress bar is no longer at the bottom but in the middle of the screen
airlied: ah then it is loading the modules too late.
EruditeHermit: also the progress bar is 3 progress bars overlayed
airlied: yup so thats the text mode startup
EruditeHermit: there is a dark blue, light blue and a white
airlied: the problem is the radeon modules isn’t in the initrd for the kerne.
spstarr: EruditeHermit: thats text mode.. not kms mode ;/
EruditeHermit: =(
EruditeHermit: airlied: the VT switch at the end of boot is smoother though
EruditeHermit: but its not like the fancy kernel mode in the video here
spstarr: EruditeHermit: does VT switch crash if you go into composite and switch back?
EruditeHermit: spstarr: didn’t try
EruditeHermit: spstarr: I’ll try next reboot
spstarr: even though that will ‘go away’ with kms taking over switching modes
EruditeHermit: airlied: is there an option to force kernel mode to work by loading something first?
spstarr: EruditeHermit: radeon.modeset=1
airlied: won’t help
spstarr: which chip is he using?
airlied: the kernel module needs to be loaded in the initrd
spstarr: oh after
airlied: not at udev time for plymouth to do its stuff
airlied: the USB images must not do that.
EruditeHermit: airlied: its the same ISO as live CD
EruditeHermit: airlied: it just uses some program to run it off USB
terracon: airlied: are you ready for Tuesday’s freeze?
airlied: terracon: its only a freeze for normal developers 🙂
EruditeHermit: lol
airlied: I don’t think we’ve ever frozen X or the kernel until 5 mins before shipping
spstarr: airlied: if there’s any consolation prize,your memory fixes in the DDX seem to have stopped the lockups in 2D w/ EXAcomposite
airlied: spstarr: with kms?
spstarr: w/o
spstarr: airlied: with kms i have these busy wait issues
airlied: wierd they should have had no effect.
spstarr: which if you can fix would give me a usable kms 🙂
EruditeHermit: airlied: so, do I just have to wait for another spin of the image?
EruditeHermit: nothing I can do to get plymouth working
airlied: EruditeHermit: well until someone fixes it, I’ve no idea how the image spins work
spstarr: EruditeHermit: what error are you having?
airlied: I’ve asked on fedora-devel so hopefully someone oknows.
spstarr: you can rebuild the initramfs
spstarr: you can tell mkinitrd to add modules
airlied: spstarr: for the licecd?
airlied: livecd?
terracon: heh lice on a cd ewwwwwwwwwwwwwwwwwww
spstarr: airlied: did he install Fedora or just booting it from live CD only
airlied: spstarr: live cd..
airlied: if he installed it would have worked
spstarr: hmm, not sure, i dont think you can load drivers that aren’t in a ramfs 🙂
EruditeHermit: so it will work if I install it?
airlied: EruditeHermit: yes should do, I probably need to play with live stuff after I get past fixing the bugs:)
EruditeHermit: ok
EruditeHermit: I’ll tell you what
EruditeHermit: I’ll dig up an old HD
EruditeHermit: and install it
EruditeHermit: let you know how it goes
EruditeHermit: does it customize the initrd when you install?
EruditeHermit: I didn’t know they did that
EruditeHermit: I thought it was all generic
airlied: Fedora still does it.
airlied: it writes out a new one on every kernel install.
EruditeHermit: cool
EruditeHermit: well
EruditeHermit: solid work so far
EruditeHermit: thanks for doing it
EruditeHermit: spstarr: I am going to restart and try again. Anything else I should try other than VT switch while using compiz?
spstarr: you have kms enabled?
spstarr: keep it enabled
EruditeHermit: spstarr: even though I don’t get graphical boot, I think I have KMS enabled VTs
EruditeHermit: spstarr: there is no flicker when it switches from the VT to X at boot
spstarr: KMS is activated on bootup when the radeon kernel module isloaded
EruditeHermit: yeah
EruditeHermit: its just not loaded fast enough
EruditeHermit: spstarr: still have the KDE4 kwin problems?
EruditeHermit: I couldn’t boot the previous snapshots
EruditeHermit: I’ll try with snapshot 3
spstarr: yes but its different now
EruditeHermit: if it is still an issue
EruditeHermit: what is wrong now?
spstarr: there are some busy wait issues airlied knows about
EruditeHermit: ok
EruditeHermit: the same as with the gnome versions?
spstarr: GNOME works with kms + composite, except not if you logout of GNOME it will cause problems
spstarr: same if you logout with KDE
EruditeHermit: spstarr: thats it?
EruditeHermit: spstarr: thats not bad
spstarr: it will also begin to busy wait if you use composite with GNOME/KDE
EruditeHermit: what is a busy/wait?
EruditeHermit: high CPU usage?
spstarr: no, lots of stalling
EruditeHermit: the system just freezes?
spstarr: pretty much off/on
EruditeHermit: I didn’t notice that with mine
EruditeHermit: but I wasn’t on for a long time
EruditeHermit: ok
EruditeHermit: well here I go
spstarr_home: airlied: i spoke too soon 🙂
spstarr_home: it crashed
EruditeHermit: spstarr: confirm all your troubles
spstarr_home: :/
EruditeHermit: spstarr: the VT switch worked for a while
EruditeHermit: then I started getting the wait times
EruditeHermit: and then the VT switch crashed it
spstarr_home: yea
spstarr_home: but thats not kms related
spstarr_home: looks at airlied 🙂
EruditeHermit: does airlied have an r300 AGP card yet?
airlied: I failed in my last effort to ebay one
EruditeHermit: is this only r300 AGP?
EruditeHermit: issue
EruditeHermit: if so, thats still good progress
spstarr_home: the busy wait issues are for the new buffmgr code
EruditeHermit: oh
spstarr_home: i think once thats fixed up those will go away
EruditeHermit: well, all problems seem to be related to the busy/wait thing
EruditeHermit: the VT thing failed after I started getting the waits
EruditeHermit: btw fedora GDM is nice
EruditeHermit: do they heavily modify it?
EruditeHermit: it has gnome power manager in there
EruditeHermit: and other cool stuff
spstarr_home: and by waits i mean, you could move the mouse.. but the desktop wold appear locked up, but then after a few mins you could interact with the desktop a little before it locked up again
EruditeHermit: yes
EruditeHermit: same here
spstarr_home: with kms on?
EruditeHermit: with kms but no plymouth rhgb
EruditeHermit: because it loaded the module too late
spstarr_home: doesnt matter about plymouth
EruditeHermit: but yes kms because the VT switching was smooth and flicker free
spstarr_home: right
EruditeHermit: Fedora uses VT1 for X
EruditeHermit: or F1 for X
EruditeHermit: debian systems its usually F7

Radeon IRC Logs For 2008-10-26

Rig Lab Hardware

You Might Also Like

RadeonHD.org | Radeon IRC Logs For 2008-10-30

RadeonHD.org | Radeon IRC Logs For 2008-11-04

RadeonHD.org | Radeon IRC Logs For 2008-11-03